llama.cpp

11f78c6a - convert-hf : adapt ArcticModel to use yield too